AI
RenderRank: Learning to Rerank Text with Compressed Visual Tokens
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 27/09/2026, 20:00:00 UTC
Briefing created: 29/09/2026, 11:20:21 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 29/09/2026, 06:18:25 UTC.
Briefings
RenderRank is a reranker that learns query-dependent relevance scoring using compressed visual document representations instead of text tokens. It achieves an average NDCG@10 of 55.96 on 11 BEIR datasets, outperforming all text-based baselines below 4B parameters and some larger models. On four long-document datasets, it reaches an average NDCG@10 of 88.27 with approximately half the input token count of evaluated text-based rerankers, delivering 1.70x the highest average throughput. The method aligns visual relevance scores with a text-based teacher, then refines relative scores for positive and negative documents.
Key points
- RenderRank uses compressed visual document representations to learn query-dependent relevance scoring, replacing text token sequences in conventional rerankers.
- On 11 BEIR datasets, RenderRank achieves an average NDCG@10 of 55.96, outperforming all text-based baselines below 4B parameters and some larger models.
Context and limitations
Results are limited to 11 BEIR datasets and four long-document datasets; specific performance metrics on other datasets are not reported.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 29/09/2026, 11:20:21 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeifjnllkts7ilbfplj2pgqgf2miwxjtsfxlyrcxgdfi7mbg4fu27om